Issue Station
Sign In
Tools hereToolbox

hermes - 💡(How to fix) Fix [Bug]: Gemma 4 MoE refuses to reason with REAP variant.

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

Error Message

Additional Logs / Traceback (optional)

Root Cause

Root Cause Analysis (optional)

Code Example

--- hermes dump ---
version:          0.10.0 (2026.4.16) [(unknown)]
os:               Linux 6.19.12-zen1-1-zen x86_64
python:           3.13.5
openai_sdk:       2.32.0
profile:          default
hermes_home:      ~/.
model:            gemma-4-21b-a4b-it-REAP-heretic.i1-Q4_K_M
provider:         custom
terminal:         local

api_keys:
  openrouter           not set
  openai               not set
  anthropic            not set
  anthropic_token      not set
  nous                 not set
  glm/zai              not set
  zai                  not set
  kimi                 not set
  minimax              not set
  deepseek             not set
  dashscope            not set
  huggingface          not set
  nvidia               not set
  ai_gateway           not set
  opencode_zen         not set
  opencode_go          not set
  kilocode             not set
  firecrawl            not set
  tavily               not set
  browserbase          not set
  fal                  not set
  elevenlabs           not set
  github               not set

features:
  toolsets:           hermes-cli
  mcp_servers:        0
  memory_provider:    built-in
  gateway:            stopped (docker (foreground))
  platforms:          telegram, homeassistant
  cron_jobs:          0
  skills:             85

config_overrides:
  agent.tool_use_enforcement: True
  compression.threshold: 0.95
  display.streaming: True
  display.show_reasoning: True
--- end dump ---

agent log contains full conversations and will not be provided for privacy reasons :/

---
RAW_BUFFERClick to expand / collapse

Bug Description

Gemma 4 REAP variant by 0xsero works fine, and reasons in every other scenario given - in llama.cpp webui, poking the endpoint myself - reasoning is present. Using the model through Hermes for some reason disables reasoning.

Steps to Reproduce

  1. Run hermes chat
  2. Optionally set reasoning to show and effort to high/xhigh (medium/low also don't reason)
  3. Ask the model anything

Expected Behavior

Model reasons, and then responds

Actual Behavior

Model doesn't reason, responds with significantly lower intelligence

Affected Component

Other

Messaging Platform (if gateway-related)

No response

Debug Report

--- hermes dump ---
version:          0.10.0 (2026.4.16) [(unknown)]
os:               Linux 6.19.12-zen1-1-zen x86_64
python:           3.13.5
openai_sdk:       2.32.0
profile:          default
hermes_home:      ~/.
model:            gemma-4-21b-a4b-it-REAP-heretic.i1-Q4_K_M
provider:         custom
terminal:         local

api_keys:
  openrouter           not set
  openai               not set
  anthropic            not set
  anthropic_token      not set
  nous                 not set
  glm/zai              not set
  zai                  not set
  kimi                 not set
  minimax              not set
  deepseek             not set
  dashscope            not set
  huggingface          not set
  nvidia               not set
  ai_gateway           not set
  opencode_zen         not set
  opencode_go          not set
  kilocode             not set
  firecrawl            not set
  tavily               not set
  browserbase          not set
  fal                  not set
  elevenlabs           not set
  github               not set

features:
  toolsets:           hermes-cli
  mcp_servers:        0
  memory_provider:    built-in
  gateway:            stopped (docker (foreground))
  platforms:          telegram, homeassistant
  cron_jobs:          0
  skills:             85

config_overrides:
  agent.tool_use_enforcement: True
  compression.threshold: 0.95
  display.streaming: True
  display.show_reasoning: True
--- end dump ---

agent log contains full conversations and will not be provided for privacy reasons :/

Operating System

Arch Linux

Python Version

docker image

Hermes Version

docker image (0.10)

Additional Logs / Traceback (optional)

Root Cause Analysis (optional)

Hermes seems to not pass down anything reasoning related down to the request if a local provider is used - high, xhigh, whatever - the parameters never actually reach the model.

Proposed Fix (optional)

Add reasoning effort to the request to local/custom providers.

Are you willing to submit a PR for this?

  • I'd like to fix this myself and submit a PR

extent analysis

TL;DR

The issue can likely be fixed by modifying the Hermes configuration to pass reasoning parameters to local providers.

Guidance

  • Verify that the display.show_reasoning config override is correctly set to True in the Hermes configuration.
  • Check the Hermes documentation to see if there are any specific settings or flags required to enable reasoning for local providers.
  • Consider adding a debug log statement in the Hermes code to confirm whether the reasoning parameters are being passed to the local provider.
  • Test the model with a different provider to see if the issue is specific to the local provider.

Example

No code snippet is provided as the issue does not contain sufficient information about the specific code changes required.

Notes

The root cause analysis suggests that Hermes is not passing reasoning parameters to local providers, but the exact fix is not clear without further investigation.

Recommendation

Apply workaround: Modify the Hermes configuration to pass reasoning parameters to local providers, as suggested in the proposed fix. This is likely to resolve the issue, but may require further testing and verification.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#ssr#memory leak#API versioning#request timeout

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING